基于聚类和AdaBoost的ADS⁃B数据质量综合评估方法  

A comprehensive evaluation method of ADS⁃B data quality based on clustering and AdaBoost

作　　者：张召悦[1] 阳颖 Zhaoyue ZHANG;Ying YANG(College of Air Traffic Management,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学空中交通管理学院,天津300300

出　　处：《航空学报》2024年第13期381-392,共12页Acta Aeronautica et Astronautica Sinica

基　　金：中央高校基本科研业务费(3122022105);天津市多元投入基金重点项目(21JCZDJC00780)。

摘　　要：为更好地发挥ADS-B数据应用价值,针对ADS-B数据质量评估过程中传统方法无法客观准确得到质量等级的问题,在分析行业应用、发射设备性能、数据安全等方面对ADS-B数据质量需求的基础上,构建了ADS-B数据质量评估指标体系,提出了基于集成学习自适应提升算法(AdaBoost)的新型数据质量评估方法。该方法通过K-means聚类确定最佳质量等级类别,结合熵权法和双基点法(TOPSIS)打分确定数据标签,并采用AdaBoost算法对评估模型进行了训练和优化。以天津机场数据为例,实验得出ADS-B数据质量的最佳等级划分为5级,得到的数据质量评估模型准确率高达98.5%,验证了该方法可以有效避免主观因素对评估结果的影响,并得到最优的质量等级划分,能够提高评估结果的稳定性和精确度。Traditional methods for ADS-B data quality assessment cannot obtain the quality level objectively and accurately.For better application of ADS-B data,an ADS-B data quality evaluation index system is constructed on the basis of analysis of ADS-B data quality requirements in industry applications,transmitting equipment performance,data security,etc.A new data quality evaluation method is proposed based on the ensemble learning Adaptive Boosting(AdaBoost)algorithm.In this method,the best quality grade category is determined by K⁃means clustering,data labels are determined by combining the Technique for Order Preference by Similarity to Idealsolution(TOPSIS),and the evaluation model is trained and optimized by the AdaBoost algorithm.The data of Tianjin Airport are use for case analysis.The experiment shows that it is the best scheme to divide ADS-B data quality into 5 grades,and the accuracy of the obtained data quality evaluation model is as high as 98.5%.This verifies that the method proposed can effectively avoid the influence of subjective factors and obtain the optimal quality grade classification,improving the stability and accuracy of evaluation results.

关 键 词：ADS-B数据质量 K-MEANS聚类 熵权法 双基点法 TOPSIS 自适应提升算法 

分 类 号：V355.1[航空宇航科学与技术—人机与环境工程]